欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
为了提高视频对象分割的性能,在传统光流法的基础上,提出了一种改进算法.该算法首先对序列图像中的帧进行预处理,平滑序列图像;然后用Gaussian金字塔降低图像的分辨率,使运动对象的速度降低;再使用结合Gaussian分布的LK光流法进行分割,得到运动对象;最后使用帧相＂与＂运算做进一步的分割.实验结果表明：该算法减少了运算时间,能更精确地分割出运动对象,取得较好的效果.
英文摘要：
This paper proposes an improved optical flow based on traditional way in order to improve the performance of video object segmentation. At first, the pre-processing method is applied to source images to smooth images. Secondly, use Gaussian pyramid reduce the resolution to make the speed of movement objects more lower. Thirdly, we segment images use Gaussian distribution with LK optical flow to get the movement objects. Finally, we apply the frame logical product make a deep segmentation. The experiment results have showed that the algorithm can decrease the operation time, accurately segment the movement object and get a good effect.
同期刊论文项目
同项目期刊论文
期刊信息
位置：